The search engine relies on the auto-generated transcripts from YouTube. The speech-to-text is far from perfect, but it's good enough to be useful.
The source material comes from Danny Roddy's YouTube channel, as well as various interviews that Ray has done over the years.
The previous incarnation of this site had a server that would handle the search requests. While this cuts down on the load time and is far more scalable, it costs money to keep the server up and would require occasional maintainence.
The current version loads all the transcripts into the browser. This creates a longer initial load, but no more server requests for search.